How much Sampling Area is Enough?: The Development of a Statistically Justified Method for Sampling Filtered Microplastics in Drinking Water

Abstract

As awareness intensifies regarding the presence of microplastics (MPs) throughout the environment, there is a growing concern that MPs in drinking water may contribute to a toxic response in consumers. With the increasingly large volumes (> 100L) being sampled to improve accuracy, yielding potentially hundreds of MPs that must be counted, coupled with the large amount of time it requires to process MP samples, there is a need to know the minimum subsample required to accurately represent the actual MP population. In order to develop a statistically justified framework for sampling, drinking water was filtered via “in-lab” and “in-line” methods. Fifteen 10 μm filters 47 mm in diameter were examined with 70 0.7 mm² grid cells, where each cell’s total MPs were counted using an optical microscope. Suspected MPs were examined in areas ranging within zero to 100% of the filter using three strategies: i) random discontiguous, ii) one random contiguous area, and iii) two random contiguous areas equidistant from the centre. Significant clustering was observed in all but two of the fifteen filters. Subsample estimates of total suspected MPs were compared with the actual MP population via consistency values. A minimum suitable sampling area was determined to exist when a specified subsample area and every sampling interval above it had a 95% confidence interval of each filter’s upper and lower confidence intervals of consistency that fell within a suitable percentage from the true value. Results of this study suggest that out of all the sampling methods tested, discontiguous sampling is optimal. Randomly sampling one quarter of the filter results in an accuracy within ± 15%, nineteen times out of twenty. Overall, it is recommended that this method be employed for future research in order to statistically justify the sampling area required for MPs in drinking water.